Skip to content
hundredfolds
beginnertutorial

Multiple Linear Regression: Predict Energy Consumption

Part 2 of the ML Regression Lab. Extend regression to several features — predict energy use from temperature and humidity — and learn to read multi-feature coefficients.

@shvinn

Machine Learning Engineer

1 min readJan 17, 2026
View source repository

In Part 1 one feature gave us a line. Real problems depend on many inputs. Part 2 generalizes to multiple linear regression: predicting energy consumption from both temperature and humidity.

Follow along: 02_energy_consumption_prediction in github.com/shvinn/ai-lab-99.

The goal

With two features the model fits a plane instead of a line:

energy=w1⋅temp+w2⋅humidity+b\text{energy} = w_1 \cdot \text{temp} + w_2 \cdot \text{humidity} + b

The workflow is identical to Part 1 — only the shape of X changes. That's the point: scikit-learn's API doesn't care how many features you have.

Step 1 — Load the data

import pandas as pd
 
df = pd.read_csv("energy_consumption_data.csv")
df.head()
df.describe()

Step 2 — Train on two features

The only difference from Part 1: X now has two columns.

from sklearn.model_selection import train_test_split
from sklearn.linear_model import LinearRegression
 
X = df[["temperature", "humidity"]]   # two features → matrix with 2 columns
y = df["energy_consumption"]
 
X_train, X_test, y_train, y_test = train_test_split(
    X, y, test_size=0.2, random_state=42
)
 
model = LinearRegression()
model.fit(X_train, y_train)

Step 3 — Interpret the coefficients

Now there's one coefficient per feature. Each is the effect of that feature holding the others fixed:

model.coef_        # → [3.41, 2.00]  per-°C and per-% effects
model.intercept_   # → 53.02

Reading it: every +1 °C adds ~3.41 kWh, every +1% humidity adds ~2.00 kWh. Because features can be on different scales, raw coefficient size isn't the same as importance — keep that in mind as models grow.

Step 4 — Evaluate

from sklearn.metrics import mean_squared_error
 
rmse_train = mean_squared_error(y_train, model.predict(X_train)) ** 0.5
rmse_test = mean_squared_error(y_test, model.predict(X_test)) ** 0.5
print(f"RMSE Train: {rmse_train:.2f}  |  RMSE Test: {rmse_test:.2f}")

Step 5 — Deploy

app.py
import streamlit as st
 
 
def predict_energy_consumption(temperature, humidity):
    return 3.40772114 * temperature + 2.00133069 * humidity + 53.02343025491177
 
 
st.title("⚡ Predict Energy Consumption")
temperature = st.number_input("Temperature (°C)", -50.0, 60.0, 25.0, 0.1)
humidity = st.number_input("Humidity (%)", 0.0, 100.0, 50.0, 0.1)
 
if st.button("Predict"):
    prediction = predict_energy_consumption(temperature, humidity)
    st.success(f"🔋 Predicted Energy Consumption: {prediction:.2f} kWh")
streamlit run app.py

What's next

So far every feature was already numeric. In Part 3 we hit our first categorical feature and learn the simplest fix: binary encoding.